Gemelo.AI的动态

I’ve been heads down. My colleagues Laurence Grant, Sapna Jayavel, Eric B. and others have taken on a challenge from #NVIDIA and Robert Kim, MBA to build accelerators for deploying AI on-premises. We did it. Our Private GPT solution is ready to launch, and it’s never been easier to try, scale, and grow this and other AI-powered solutions. Here are a few insights we’ve gathered along the way: ? NVIDIA’s AI Enterprise (AI-E) platform simplifies the deployment of AI models, whether you’re working with industry AI/Deep Learning models or GenAI models like LLAMA3. ? The stack is designed to follow native app/container development and orchestration approaches, making deployments and scaling straightforward. ? NVIDIA has done the heavy lifting of optimization through their NIM packaging structures. ? You can bring any open-source, developed, or commercial solutions for RAG (aka, Chat with Your Data), Agentic-AI (Agents/Workflow/Bots), Creation (multi-model), Doc-Processing (NER) and more. Or start with #Presidio’s Accelerator for Enterprise RAG. ? The same Python packages that work with OpenAI work seamlessly here. My teammates have put in countless late nights mastering the #NVIDIA AI Enterprise platform, the #NIM architecture, and orchestration. We’ve partnered with #DELL, #CISCO, #HPE, and #NVIDIA's native #DGX kit to create POC, PILOT, and PROD scale versions. As always, feel free to reach out. Let’s build together. Thanks for all your support, Shae Fogg (Kirkman)!

#AI #Bubble #AIHype #GenerativeAI #OpenAI #Nvidia: "In September 2023, I published?AI’s $200B Question. The goal of the piece was to ask the question: “Where is all the revenue?” At that time, I noticed a big gap between the revenue expectations implied by the AI infrastructure build-out, and actual revenue growth in the AI ecosystem, which is also a proxy for end-user value. I described this as a “$125B hole that needs to be filled?for each year of CapEx at today’s levels.” This week, Nvidia completed its ascent to become the most valuable company in the world. In the weeks leading up to this, I’ve received numerous requests for the updated math behind my analysis. Has AI’s $200B question been solved, or exacerbated? If you run this analysis again today, here are the results you get: AI’s $200B question is now AI’s $600B question." https://lnkd.in/dJV-spHC

Ever thought of having a supercomputer on your desk? ?? NVIDIA? Project DIGITS, a personal AI supercomputer that brings the power of the NVIDIA Grace Blackwell platform to your fingertips! ?? Grace Blackwell Superchip: This powerhouse delivers a whopping petaflop of AI performance! ?? Run Massive Models: Train and run complex AI models with up to 200 billion parameters – right on your desktop. ?? Seamless Deployment: Develop and test models locally, then seamlessly deploy them to the cloud or data center. Project DIGITS empowers researchers, data scientists, and students to develop cutting-edge AI at their desks. #nvidia #digits #ai #machinelearning #deeplearning #supercomputer #science #research

“If (you’re) waving the white flag and saying ‘Well, not fair! We want to still sell (colocation) like we did 10 or 20 years ago.' The interesting part with the ‘not fair’? When you talk to someone like Jensen Huang for NVIDIA, this industry sees a trillion dollars coming into data centers in the next four years… And when you talk to Michael Dell, who sees 100-fold growth in data centers over the next decade… We're all profiting off of AI, right? Generative AI going into data centers, high-performance computing (HPC). All this stuff is a massive growth enabler for the data center industry. Yet, we're sitting here and saying, ‘Well, how about we sell that (colocation) out of the back of our Model T Ford? Or our horse and buggy?' (no wait, I'll fax it to you :-) Because our prospects won't notice that everyone else in the world has websites with lead-generation content, videos, webinars, and all that stuff. We'll just pretend that we're so special that we can keep marketing and selling like we did 20 years ago, and they won't notice. Like, seriously, that's your playbook?” See comments for the link to the full episode #datacentergtm #dcsmi #DataCenter ?? #AI ?? #FutureTech ?? #DigitalTransformation ?? #AdaptOrDie ?? #InnovateOrStagnate ??

The DeepSeek AI Disruption: Who Wins in the AI Infrastructure Race? #DeepSeek’s open-source LLM is shaking up the AI landscape - not just for Big Tech but also for SMBs that were previously locked out of the generative AI race. While some feared a negative impact on NVIDIA (#NVDA) due to DeepSeek’s cost-efficient approach, the reality looks different. The demand for GPUs and AI infrastructure isn't declining - it’s diversifying. This shift benefits “shovel sellers” - the infrastructure providers enabling LLM builders and AI app developers, regardless of which model they use. With competition heating up, providers will battle for clients by lowering subscription fees, increasing token usage, and offering new services. But no matter which #LLMs dominate, the growing need for AI compute power remains undeniable. The Total Addressable Market (TAM) is massive: ?? AI Data Management → $106B+ (Grand View Research) ?? Autonomous Driving Tech → $315B (Frost & Sullivan) From GPUs-as-a-service to AI-driven data solutions, edtech, and more, infrastructure players are well-positioned to thrive. OpenAI once defined what was possible, but with #Stargate’s high-cost approach contrasting DeepSeek’s affordability, the #AI ecosystem is evolving fast. The real winners? The ones selling the picks and shovels in this AI gold rush. #AI #DeepSeek #LLMs #Nvidia #GenerativeAI #Infrastructure

?? Breaking Boundaries in AI Computing! NVIDIA has just unveiled Project DIGITS, a personal AI supercomputer that’s bringing the power of high-performance computing right to your desk. ????? ?? Imagine running large language models with up to 200 billion parameters, all locally, without relying on cloud resources. With the new GB10 Grace Blackwell Superchip, delivering 1 petaflop of AI performance, this compact marvel empowers AI developers, researchers, and even students to innovate faster than ever. ? In my latest Substack post, I explore: ?? What is NVIDIA’s Project DIGITS? ?? How it will transform AI development. ?? The unprecedented power it brings to your fingertips. If you’re as excited about the future of AI as I am, check out the blog here: https://lnkd.in/gg_BskjF Let’s discuss how this could shape the next wave of AI innovation! Share your thoughts in the comments. ???? #ArtificialIntelligence #NVIDIA #ProjectDigits #AIRevolution #TechInnovation #MachineLearning #DataScience #Supercomputing #AIResearch #FutureOfTech

?? DeepSeek wiped out over $600 billion this week from AI market leaders combined market value. ?? Is this a reasonable reaction to a new startup without validating the facts? ?? Think about it - DeepSeek has released their model weights and some components under their license, but hasn’t shared critical elements like training data, complete source code, and reinforcement learning implementation details. This makes it more accurately described as ‘open weights’ rather than fully open source, which is still significant but different from traditional open-source software. As Sam Altman and others have said in the past 3 days - it is easy to stand on the shoulders of those ahead of you and claim a victory. Waiting for someone from the research community (including the Hugging face team) to reproduce the magic at their end to confirm how much of this is real and how much is just smokes and mirrors. Till then, enjoy the memes and how to videos on DeepSeek - NVidia GPUs hard at work ?? ?? Subscribe to my newsletter and share this post: https://lnkd.in/g3bdneR7 #ai #genai #aiadoption #techhumor

AllCloud is proud to partner with #AWS and QuesTek Innovations LLC to revolutionize materials innovation with #generativeAI. ???? ?? Read about how we helped accelerate this breakthrough: https://go.aws/3VpaBr1 #MaterialsScience #AI

AI is becoming more stateful vs. stateless Being able to read/listen, understand, and then accurately remember or respond with the right historical context is key to wider GenAI adoption The number of times I hear "Honey, I can't believe you forgot xyz!" is something yours truly is working to reduce as I get older and life gets ever more busy ?? One takeaway as the GenAI industry matures is that a stateful and "memory-rich" AI architecture matters a great deal in being able to perform well in complex tasks and deeper interactions with real people The concept of memory checkpointing plays an increasingly important role in both model training and inferencing I'm very excited about the role MemVerge is playing in the industry in partnership with leaders like NVIDIA to maximize GPU infra utilization and improve AI model development and inferencing Check out this podcast interview of our very own Bernie Wu who provides a great lay of the land in terms of memory architecture and new innovations as it pertains to K8s and AI (URL provided in the comments) #GenAI #AWS #NVIDIA #AI #HPC

Exciting News! ?? Deepchecks has been featured in NVIDIA's partner blog about NVIDIA AI Foundry and custom generative AI models. NVIDIA AI Foundry is a service that empowers enterprises to leverage data, accelerated computing, and software tools to create and deploy custom generative AI models.?? Users can develop NVIDIA AI Foundry models for production using AIOps and MLOps platforms from NVIDIA partners, including Deepchecks and more. ?? Learn more here: https://lnkd.in/dPkzPUGu #GenAI #NVIDIA #LLMs #LLMOps #AIOps #MLOps